Control point adaptation algorithm for CAD-based visual tracking of micro MEMS parts using active stereo vision system

Visually guided assembly of micro MEMS parts with high magnification is limited by small field of view (FOV). To overcome this problem, zoom lens system is usually adopted, which causes size change of the parts in the image space. Micro MEMS parts have several arbitrary planar shapes because of the fabrication process, so that contour to be described by control point can be suitable image feature to recognize the object and estimate its pose. In this case, many control points can describe the shape in detail, but they cause the computational problem. It is important to visual tracking or visual servoing which need to compute the algorithm in real-time. Also, the object size variation in image plane by zoom needs to change the number of control points in order to describe object. This paper proposes a new stereo vision system with zoom mechanism and control point adaptation algorithm. The proposed vision system is equipped with a top camera and a side camera with a rotating mirror, which obtain images using zoom lens systems. By control point adaptation, the number and placement of the control points are adjusted to efficiently describe the shape of the object. Arbitrary shaped planar MEMS parts are used to verify the effectiveness of the proposed algorithm. Experiment results show the feasibility of the proposed algorithm for assembly of micro MEMS parts.